视线估计方法、系统、电子设备、存储介质、程序产品
By using a prediction model trained with multiple consecutive frames in the gaze estimation method, combined with regression and classification loss functions, the accuracy and robustness of gaze estimation are significantly improved. This solves the problem of insufficient gaze estimation in different environments in existing technologies and is applicable to driver detection and autonomous driving assistance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JILUO TECH (SHANGHAI) CO LTD
- Filing Date
- 2022-04-27
- Publication Date
- 2026-07-17
AI Technical Summary
Existing gaze estimation methods are not robust to different environments and are difficult to use effectively in real-world situations, especially in driver detection systems. The models perform well on a single public dataset but are not accurate enough on diverse datasets.
When training the prediction model, a series of labeled consecutive frames are input, and the model is trained using a combined loss function of regression and classification loss functions. The gaze estimation is performed using the time-series information of the multiple frames, and the combined result of gaze angle and event is output.
It significantly improves the accuracy and robustness of gaze estimation, enabling the model to predict gaze direction more accurately in different environments, and is suitable for driver detection and autonomous driving assistance.
Smart Images
Figure CN114898255B_ABST